Limits and Alternatives to Multiple Regression in Comparative Research

This paper criticizes the use of multiple regression (MR) in the fields of comparative social policy and political economy and proposes alternative methods of numerical analysis. The limitations of MR in its characteristic guise as a means of hypothesis-testing are well known. The emphasis here is on the specific difficulties of applying MR to the problem of explaining diverse outcomes across a limited range of country cases. Two principal conclusions will emerge. First, even though technical means are available to deal with many of the limitations of MR, these solutions are either unconvincing or else require such advanced technical skills that they offer questionable returns on scholarly investment. Second, dissatisfaction with MR does not necessarily mandate radical alternatives or abandonment of numerical methods altogether. ‘‘Low-tech’’ forms of analysis (tabular and graphical methods) and multivariate statistical techniques other than MR (such as factor analysis) constitute viable and useful alternatives. The comparative study of welfare states is a good example of the characteristic methodological polarization that afflicts the social sciences. Historians and social policy analysts with an intrinsic interest in welfare states engage in descriptive and prescriptive studies, while at the other extreme are ‘‘hard-nosed’’ social scientists who regard the welfare state essentially as a